of the fundamental truths about credit score models is that they must
periodically be refreshed in order to maintain a peak level of
predictiveness. Not only does the data available to modelers get better
and model building techniques improve, but the mix of credit products
changes, as do consumer behaviors and the way they treat their financial
For example, let’s take a look at the massive
increase in student debt. Models built prior to the Recession didn’t
factor in how consumers with large student debt obligations would
influence consumer credit behavior.
For this reason, our data
scientists took a deep dive into the credit usage patterns of different
generations and they layered on extra factors indicative of financial
health (e.g., income and assets). By doing this, we found that there may
be a paradigm shift that methodologies underpinning older scoring
models are not considering.
Included in this newsletter is a deeper dive, but I’ll topline some of the important findings:
speaking, the assumption was that those with higher income and assets
were associated with thicker credit files and more credit usage (i.e.,
thick file consumers have three or more credit accounts reported and
thin file consumers have two or fewer credit accounts). Relatedly,
lenders often viewed thin file consumers as more risky than those with
thick files and they are often placed into the highest risk products
(think: high interest rates and modest loan limits).
what we are now seeing is that Millennials with thin files – unlike any
other generations before them – on average have income and asset levels
consistent with their thick file counterparts.
this, conventional models and lending strategies might actually be
penalizing them simply because that’s historically how thin file
consumers have been treated.
- Accordingly, users of credit
scores for lending decisions should carefully assess whether they should
reconsider models based on legacy beliefs.
This is an
opportunity for lenders to lean into the Millennial generation and get a
firmer understanding of how they are handling their credit health. From
a credit scoring standpoint, it also speaks to the importance of
trended credit data.
Here’s why: trended credit data examines
the longer term trajectory of credit behaviors as opposed to a snapshot
or single point in time from the prior month. A model that uses trended
credit data (like VantageScore 4.0) is better able to understand more
recent credit behaviors and relies less on some of the more conventional
attributes used by models that focused on the tenure, breadth and depth
of credit usage – which, by definition, negatively impacts those who
are new to credit.
In other words, the richness of trended
credit data allows our data scientists to extrapolate predictive
behaviors from consumers who choose to open less credit accounts.
share more of these types of insights in the coming weeks and months.
For now, we’ve highlighted some insights here, but I also encourage you
to follow us on LinkedIn where we can continue the conversation.
in prior years, we won’t be putting out a December newsletter so this
will be the last one for 2018. It was a great year for VantageScore in
our 12th year of continuous growth, and we thank you for supporting our mission and reading our newsletter.
Happy holidays to all,
CEO and President, VantageScore Solutions